Does CSF testing add value after an indeterminate plasma p-tau217 result?

Background It is unclear whether cerebrospinal fluid (CSF) testing improves classification when plasma phosphorylated tau 217 (p-tau217) results are indeterminate. Using amyloid and tau positron emission tomography (PET) as the reference standard, we compared plasma p-tau217 with core CSF biomarkers and evaluated a sequential plasma-to-CSF workflow. Methods We studied 283 Alzheimer's Disease Neuroimaging Initiative participants (148 cognitively unimpaired [CU]; 135 symptomatic) with concurrent plasma p-tau217, CSF biomarkers and amyloid PET; 265 also had tau PET. Amyloid positivity was defined at 18, 37 and 45 Centiloids (CL). Areas under the receiver operating characteristic curve (AUC) were compared with paired DeLong tests and false discovery rate (FDR) correction. Indeterminate plasma results (between the 95%-sensitivity and 95%-specificity thresholds) were resolved with internally optimized CSF p-tau181/Aβ42 or Aβ42/40 cutoffs; incremental value was assessed with net reclassification improvement (NRI) and McNemar tests. Results Mean (SD) age was 71 (7) years; 153 (54%) were women. No AUC difference between plasma p-tau217 and either CSF biomarker survived FDR correction, in either group, at any CL threshold or Braak stage. Internally optimal plasma p-tau217 thresholds shifted with clinical status and with the target amyloid and tau burden. Sequential CSF testing did not significantly improve classification (amyloid NRI 0.01 to 0.10; tau NRI − 0.09 to 0.08) and misclassified up to 27% (amyloid) and 33% (tau) of indeterminate cases. Conclusions These data are compatible with, but do not establish, equivalence between plasma p-tau217 and CSF biomarkers. When plasma results were indeterminate, CSF testing did not reduce uncertainty in this cohort. Clinical trial number: not applicable.